Methodology
| Data | Methodology |
|---|
Literature Review
Prior to the application of the Huff’s model as well as the MCI (multiplicative competitive interaction) model in our project, our group decided to explore the contexts with which those models have been applied elsewhere in the real world. This is so that we can understand the models better and also figure if we can learn anything which could be applied in the context of our project eventually.
Huff’s model which was initially discovered by David Huff in 1962 (Huff D.L, 1962) stated that the probability of any individual choosing a store is a ratio of the utility of that store to the sum of utilities of all other stores that the individual considers. In the context of the library, the utility/attractiveness of a store was determined by the size of the library and the transportation costs involved.
Huff’s basic model was then modified to attempt to include additional determinants of attractiveness for a conclusive picture of utility of a retail area. (Thomas, 1976) Reputation of the library would be one such example of a determinant of attractiveness.
The MCI model eventually got introduced. It extended the Huff’s Model and basically stated that attractiveness should capture the essence of competitive interactions. (Nakanishi & Cooper, 1974). In the context of the library, this meant that some of the factors included in the attractiveness index could be considered as substitutes for the library and can eventually cannibalise patrons from going to the public libraries. Examples of such factors would include bookstores and other entertainment facilities.
A further extension of the model showed how both external and internal factors of attractiveness should be considered (Jain and Mahajan, 1979). This was eventually used in a project in Germany which determined the agricultural mix of the farming sector i.e. whether farmers should go into cash crops, dairy production etc. (Neuenfelt, 2014). So other than external factors of attractiveness such as the socioeconomic factors, internal factors dealing with the farms’ capabilities were also considered (land, labour, capital etc.). In the context of our project, this would mean looking at the internal facilities of the library such as the availability of study spaces or a café.
Revision of Methodology
Adaptation of the Multiplicative Competitive Interaction Model
Based on multiple sessions with our supervisor, our team has decided to use the MCI package in R to conduct the analysis of attractiveness using Huff’s model. The Multiplicative Competitive Interaction (MCI) Model (Nakanishi/Cooper 1974), is an “econometric model for analyzing market shares in a competitive environment where the market is divided in i submarkets (e.g. groups of customers, time periods or geographical regions) and served by j suppliers (e.g. firms, brands or locations)”. Resulting market share of the suppliers (libraries in our case), this model also analyzes the attraction/utility of the alternatives in the submarket.
Different from the senior group’s model, MCI model is nonlinear but can be transformed via Ordinary Least Square (OLS) regression using the multi-step log-centering transformation and our team will also re-arrange the raw data in an interaction matrix to fix into the model.
The purpose of this is to first use the MCI model along with the different new variables for attractiveness to assign weights to the variables. This would tell us which variables actually contribute towards the attractiveness of a library. After which we would then use the Huff’s model function in the MCI package to give us the probability (𝑷_𝒊𝒋) that a Patron from a given Subzone i would visit a given Library j. One of the main motivators of shifting towards the MCI model is so that there is no longer a need to store the hard-coded Huff’s model calibrations within the dashboard, which is how the previous team had kept their results. With the implementation of the MCI model, the users of the dashboard can then recalibrate the model within the dashboard to produce new Huff’s model attractiveness based on the buffer stated by the users.
All in all, based on our literature review, we are looking into including the following factors into our analysis at the moment. (They will be subject to changes depending on the progress of the project and the practicality of applying these factors):
- 1.Distance from Subzones/Planning Area to Library
- 2.Number of MRTs present within a Buffer
- 3.Number of Malls present within a Buffer
- 4.Number of Tuition Centres within a Buffer
- 5.Collection Size of the library
- 6.Gross Floor Area of the Library
- 7.Carpark accessibility
- 8.Branch Type (Mall, Stand-Alone, Regional)
- 9.MRT Centrality
